grouping likert items

[PeteM]

First my apologies if this has already been covered, I have searched but been unable to find anything on this specific point.

the objective of my research is to investigate any difference in the Weltanschauungen of actors in the resource consent process and the effect this has on their perception of process and outcomes. I have a 60 statement questionnaire using a 5 point Likert scale.

My question is, is it statistically valid to combine the responses for say 4 statements and analyse these as representing the respondents views on substitution of natural capital. My reading of Lkert's paper is that this is valid and represent a 'Likert scale' of four 'Likert items'. Is this correct? I know the analysis of Likert scale data is a minefield of conflicting opinions, so any suggestions of papers that would support this technique would be most appreciated. BTW i was intending to do the analysis using Kruskal Wallis as I really can't see data from a Likert item as interval.

cheers

Pete

 

[spunky]

i think i'd somewhat venture the guess that in order for you to group likert-type items as to claim they represent something else you'd need to do some sort of factor analysis on them... however a little bit more information on your research hypothesis would be helpful. i'm not sure i quite understand what you're trying to do...

 

[Csionline]

You can definitely "group" likert items, it is called creating a composite. However, you must make sure that the items are (1) theoretically related and (2) significantly correlated. Also, you want to make sure the scale of each item is the same otherwise you have to standardize all the items first.

jason

http://thecsionline.org